We use a combination of APIs from existing LLM providers, and do some serious prompt engineering to get the best from them. We're starting to train models on mo

We use a combination of APIs from existing LLM providers, and do some serious prompt engineering to get the best from them. We're starting to train models on more SQL-specific prompts now.